Stereo Crosstalk Elimination (reduction) Par Excellence!

[Lion♡]

Sorry if I'm missing the point, but I though it was only applied to the opposite ear! It just has to be done over and over again because both ears pick up the correcting signal, unless a local wearable device was used right at each ear to correct the crosstalk. But in this case, if it worked perfectly, then still the opposite ear would hear nothing from that channel, so it'd just be simpler because the correction signal would not have to be recursive. I still think we have to find a way to reduce the effect as sounds are panned further left or right.

Ideally, that would be the case, but in reality, it doesn’t work that way.
The sound comes from the left speaker, and the left ear hears it first. Then, after the ITD, the right ear hears it.
At this point, the right speaker sends out a cancellation signal for it. While this signal cancels the sound for the right ear, the left ear also hears the cancellation signal.
That’s why repeated signals are cross-fed to erase unintended traces.
The graph I shared shows the results of applying a filter that specifically targets only the opposite ear. (Of course, this isn’t possible in reality.)

similar to what Dolby Pro-Logic up-mixing does.

This is a bit off-topic, but I have almost implemented Dolby Pro Logic IIX using EQ APO (excluding dynamic steering, as EQ APO cannot utilize real-time level detection to assign weights, so it would need to be re-created in Reaper).
While working on this, I also started studying Logic7, which many people seemed to appreciate.
I recall reading some documents about the Lexicon DC-1 and noticing that it also had a form of crosstalk cancellation hidden within it.

 

[Tim Link]

Ideally, that would be the case, but in reality, it doesn’t work that way.
The sound comes from the left speaker, and the left ear hears it first. Then, after the ITD, the right ear hears it.
At this point, the right speaker sends out a cancellation signal for it. While this signal cancels the sound for the right ear, the left ear also hears the cancellation signal.
That’s why repeated signals are cross-fed to erase unintended traces.
The graph I shared shows the results of applying a filter that specifically targets only the opposite ear. (Of course, this isn’t possible in reality.)


This is a bit off-topic, but I have almost implemented Dolby Pro Logic IIX using EQ APO (excluding dynamic steering, as EQ APO cannot utilize real-time level detection to assign weights, so it would need to be re-created in Reaper).
While working on this, I also started studying Logic7, which many people seemed to appreciate.
I recall reading some documents about the Lexicon DC-1 and noticing that it also had a form of crosstalk cancellation hidden within it.

View attachment 419993

That angular dependence issue is something I've played with, so that's also something that could be much more easily worked on with STC, maximizing the crosstalk below 700 Hz. This can also be done by placing drivers further apart at lower frequencies. But then there are new perceptual issues that arise. I have read that people have tried using wider low frequency spacing along with STC for each speaker spacing, and reported excellent results. For just walking around the room listening, it's best to have location coherency down to at least 100 Hz. Probably 80 Hz.

 

[Lion♡]

That angular dependence issue is something I've played with, so that's also something that could be much more easily worked on with STC, maximizing the crosstalk below 700 Hz. This can also be done by placing drivers further apart at lower frequencies. But then there are new perceptual issues that arise. I have read that people have tried using wider low frequency spacing along with STC for each speaker spacing, and reported excellent results. For just walking around the room listening, it's best to have location coherency down to at least 100 Hz. Probably 80 Hz.

The steering I mentioned refers to the steering found in upmixers.
Rather than using fixed angles, it extracts the volume differences and directional information from the L and R channels of the audio source. Based on this, the mixing coefficients dynamically change, allowing real-time adjustments to the weights and altering the angles accordingly.

In EQ APO, this is not possible, as only fixed angles can be used. Logic7, in particular, seemed to make more active use of such rear-angle processing.

Additionally, utilizing decorrelation below 700 Hz could provide quite an interesting experience.
I apply this in my custom upmixing code and artificial reverb as well, so feel free to give it a try. I'll send you a PM.

 

[Punchrulle]

https://www.embracingsound.com/Old/docs/EmbracingSoundTechSummary.pdf

SALORA orthoperspecta

(Google translate is your friend)

MULTI-CHANNEL AUDIO

 

[Tim Link]

https://www.embracingsound.com/Old/docs/EmbracingSoundTechSummary.pdf

SALORA orthoperspecta

(Google translate is your friend)

MULTI-CHANNEL AUDIO

Interesting information! The Salora system I feel I understand completely. I've tried that. It ain't bad.
Whatever EmbracingSound is doing I only partially understand. It seems they are doing almost what I was doing except somehow incorporating the L+R into the two speakers instead of using a stand-alone center, and then doing some magic of some sort to get the frequency response to be pretty even. L - R + L + R just gives 2L except for their magic circuit is somehow doing something of a comparison to make something else happen. My guess is they are electronically deriving a center channel, playing that through both speakers, and then whatever is left over for L and R is sent to each speaker separately but out of phase. The divider plate then makes sense because the drivers will try to cancel themselves out otherwise. And also they apply some kind of dyname EQ to make sure the overall energy level at each frequency remains consistant with what was in the recording. I'd be interested in hearing it. It's obviously still going to have HRTF anomalies due to the fact that sounds seem to be coming form directions where there isn't actually a speaker.
At this point, after a lot of listening and experimenting, I'm leaning more in the direction of high quality upmixing. I'm less interested in hyper envelopment and more in imaging clarity and coherence. Most upmixers focus on surround sound and don't give me the flexibility to decide just how I want the sound to be steered. I'm now studying DSP so I can learn to make my own upmixers and try to make something I may find more satisfying. There are so many good two channel recordings that I'd like to hear pretty much the same as I'm hearing them in regular stereo, but without so much phantom image comb filtering. 7 channels spread out in the normal 60 degree listening triangle is something I'd like to hear. A normal width soundstage with much more audibly coherent direction cues across it's width.

 

[Punchrulle]

Interesting information! The Salora system I feel I understand completely. I've tried that. It ain't bad.
Whatever EmbracingSound is doing I only partially understand. It seems they are doing almost what I was doing except somehow incorporating the L+R into the two speakers instead of using a stand-alone center, and then doing some magic of some sort to get the frequency response to be pretty even. L - R + L + R just gives 2L except for their magic circuit is somehow doing something of a comparison to make something else happen. My guess is they are electronically deriving a center channel, playing that through both speakers, and then whatever is left over for L and R is sent to each speaker separately but out of phase. The divider plate then makes sense because the drivers will try to cancel themselves out otherwise. And also they apply some kind of dyname EQ to make sure the overall energy level at each frequency remains consistant with what was in the recording. I'd be interested in hearing it. It's obviously still going to have HRTF anomalies due to the fact that sounds seem to be coming form directions where there isn't actually a speaker.
At this point, after a lot of listening and experimenting, I'm leaning more in the direction of high quality upmixing. I'm less interested in hyper envelopment and more in imaging clarity and coherence. Most upmixers focus on surround sound and don't give me the flexibility to decide just how I want the sound to be steered. I'm now studying DSP so I can learn to make my own upmixers and try to make something I may find more satisfying. There are so many good two channel recordings that I'd like to hear pretty much the same as I'm hearing them in regular stereo, but without so much phantom image comb filtering. 7 channels spread out in the normal 60 degree listening triangle is something I'd like to hear. A normal width soundstage with much more audibly coherent direction cues across it's width.

”My guess is they are electronically deriving a center channel, playing that through both speakers, and then whatever is left over for L and R is sent to each speaker separately but out of phase. The divider plate then makes sense because the drivers will try to cancel themselves out otherwise.”

That is my understanding of it to.
I actually have a EmbracingSound speaker (and the active ese-1 box that do the EmbracingSound magic).
It’s quite good, especially with movies.

 

[tmuikku]

ps. Salora Ortoperspekta is Finnish product, developed by by Tapio Köykkä. There is much talk and enthusiasm around this old concept, even some later variations in Finnish forum: https://foorumi.hifiharrastajat.org/index.php?search/251259/&q=ortoperspekta&o=date

Some old Salora systems for sale every now and then.

Salora Ortoperspekta | Tori

Kaikki osat tallessa, ollut päivittäisessä käytössä vielä pari vuotta sitten ja sen jälkeen säilytetty sisätiloissa. Kuvissa hieman jotain irtoroskaa. SALORA HIFI STEREO 3000:n stereojärjestelmä kolmikanavainen ns. orthoperspekta. TPerusajatuksena on välittää tila kaksikanavaista stereota...

www.tori.fi

 

[Punchrulle]

The ortoperspekta set up is very apealing to me. My take on it will be EmbracingSound center (stereo) with side DML panels (300-3000hz).
It Will differ from the ortoperspekta concept but it has some similarities. We’ll see how it works with DML sides.

 

[Heinrich]

See also:

Stereo recording techniques and setups

Learn all about mic spacing and angling in order to setup recording stereo techniques like AB, XY, near coincident, ORTF, DIN, NOS and more!

www.dpamicrophones.com

From that you may conclude, that 'stereo' is not equal 'stereo'. In other words, the stereo on your recordings is a made-up thing that only weakly correlates to what real ears would hear. The people at the mixing console take the input from the microphone arrangement as raw material and, as you expect, mix it together.

The purpose of the mixing process is to generate a satisfying impression - when listening to a playback in the commanded arrangement of the speakers, the infamous stereo-triangle. W/o any alterations to either sound wave propagation or up/down-mixing channels.

That's why elder and any new approach to bettern the stereo image by manipulation is deemed to be void.

Conversely, you shall do as the mixer expects you to do. Not the real scene is the real thing, but what the mixer hears. All reality of the stereo is in the mix, it has not real foundation, it's a simulation. The simulation is tested first by the mixer's ears, and if you don't trust the initial verdict, well, then you're lost in translation.

 

[Tim Link]

See also:

Stereo recording techniques and setups

Learn all about mic spacing and angling in order to setup recording stereo techniques like AB, XY, near coincident, ORTF, DIN, NOS and more!

www.dpamicrophones.com

From that you may conclude, that 'stereo' is not equal 'stereo'. In other words, the stereo on your recordings is a made-up thing that only weakly correlates to what real ears would hear. The people at the mixing console take the input from the microphone arrangement as raw material and, as you expect, mix it together.

The purpose of the mixing process is to generate a satisfying impression - when listening to a playback in the commanded arrangement of the speakers, the infamous stereo-triangle. W/o any alterations to either sound wave propagation or up/down-mixing channels.

That's why elder and any new approach to bettern the stereo image by manipulation is deemed to be void.

Conversely, you shall do as the mixer expects you to do. Not the real scene is the real thing, but what the mixer hears. All reality of the stereo is in the mix, it has not real foundation, it's a simulation. The simulation is tested first by the mixer's ears, and if you don't trust the initial verdict, well, then you're lost in translation.

I talk to people who mix and master and find that they have learned not to have very specific expectations about how people are going to be listening. Headphones? Small Bluetooth speaker? Car speakers? Standard 60 degree triangle on high end towers? They just don't know, and they try their mixes in a variety of environments and make the best compromises they can.
It is interesting to listen to a variety of recording styles and find they can be satisfying in various ways depending on how you are set up. Certain speaker arrangements are going to have audible deficiencies and there's nothing the mixer/recorder can do about it. Sometimes it may sound better to do something different, depending on the content and your mood. I was experimenting recently with what different recordings sound like when one channel has it's polarity reversed. For those that are recorded mono from stems and then volume panned, the imaging goes out the window when this done. However, with older Mercury Living Presence recordings that had spaced apart microphones at a distance from the orchestra, the effect is much less noticeable. The instruments pretty much stay in their same locations on the sound stage. The phase is already misaligned so it doesn't matter so much. There should be a noticeable effect in the middle because the center mic. signal, which is fed in to both channels equally, should be directly out of phase with itself. I think I do hear it, but the imaging is still pretty similar.
I did that test because I was wondering what a derived center channel algorithm might do with such a recording. If it's looking for correlation, it might have a hard time finding much of it with that recording, except for the center microphone feed. It should pull that signal neatly into the center channel and pretty much leave everything else alone. It makes me want to try it!